AMD Focuses on Bringing AI to Edge Devices, Especially CPUs

While companies like NVIDIA are trying to build GPUs and bring AI supercomputers into the tech world, AMD seems to be more focused on bringing AI to edge devices, especially CPUs.

"What I want to emphasize is that AMD's strategy in AI is much broader than just focusing on GPUs. That's what sets us apart from NVIDIA," said Gilles Garcia, Senior Director of AMD's Data Center Communications Group, at the IMC 2023 event in New Delhi.

AMD believes that most AI workloads can be handled by CPUs alone. "CPUs are best suited to handle most of the challenges faced by current edge processing, such as thermal management, cost-effectiveness, and reducing footprint by working on edge devices," Garcia said.

He emphasized that by using only CPUs as the computational engine for AI processing, AMD's large number of cores can handle a significant workload. This is made possible by AMD's Siena, which was launched in 2022. He added, "The industry has been able to offer 32 cores within a power range of about 200 watts. With Siena, AMD is able to offer 64 Zen 4 processor cores at the same power consumption."

"If you want to use GPUs on edge devices, you may face challenges in terms of latency and power consumption. So in this case, you may need to look for alternative solutions other than GPUs that can reduce power consumption without using external GPUs," Garcia said. "Depending on the workload, CPUs can handle AI workloads on their own. Of course, you may still need accelerators in the future, and that's where GPUs come in."

Therefore, AMD is not solely focused on a single strategy but is paying attention to various areas of AI without sparing any effort. In June of this year, AMD CEO Lisa Su announced plans to soon launch the Instinct MI300X to compete with NVIDIA's H100 product. To this end, the company is also focused on further developing ROCm before the hardware release.

Clearly, AMD has expanded its scope beyond GPUs and AI and is now actively involved in various technology fields, with a particular focus on edge computing and workload optimization.